Tunnel strong earthquake belt lining cavity scanning system of three-dimensional laser radar

Abstract

This invention belongs to the field of tunnel cavity detection technology, and provides a three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones. The system includes: a tunnel contour point cloud data acquisition module for collecting tunnel contour point cloud data; a cooperative positioning data acquisition module for collecting cooperative positioning data; a data fusion module for fusing multi-period tunnel contour point cloud data and cooperative positioning data; a standard tunnel contour model construction module for processing the fused data using radial basis function neural networks and clustering algorithms to establish a standard tunnel contour model; and a lining cavity detection module for identifying and detecting lining cavities based on the standard tunnel contour model and measured tunnel contour point cloud data. This invention enables high-precision and high-efficiency scanning and detection of tunnel lining cavities in strong earthquake zones, improving the accuracy and reliability of detection.

Description

Technical Field

[0001] This invention relates to the field of tunnel cavity detection technology, and in particular to a three-dimensional lidar system for scanning cavities in tunnel linings in areas of severe earthquakes. Background Technology

[0002] With the acceleration of urbanization and the rapid development of underground transportation networks, subway tunnels, as an important component of urban infrastructure, have attracted much attention regarding their structural safety and long-term stability. However, during long-term operation, tunnel structures are susceptible to geological activity, external vibrations (such as earthquakes and traffic vibrations), and material aging, leading to potential risks such as the propagation of micro-cracks and the formation of macro-voids. Failure to detect and assess these risks in a timely manner may trigger serious accidents such as localized collapses, water seepage, or even structural failure, threatening operational safety and the safety of public life and property. Currently, tunnel structural health monitoring mainly relies on the following technologies, but significant limitations remain: Traditional detection methods depend on visual inspection by technicians or measurement with handheld devices, resulting in low efficiency, limited coverage, and strong subjectivity, and are difficult to capture dynamic changes in real time; Single-sensor monitoring, relying on lidar scanning to obtain high-precision three-dimensional deformation data, lacks the ability to detect the depth of internal cracks and cavities; vibration sensors monitor the intensity of external vibrations, but this is not effectively combined with structural damage models, making it impossible to quantify the impact of vibration on tunnel safety; In terms of data processing, static risk assessment is only carried out through finite element simulation or empirical formulas, without considering the time-varying characteristics of cracks and cavities, resulting in predictions lagging behind the actual risk development; Data from different sensors are processed independently, failing to achieve multi-source information fusion, leading to biased risk assessments; Tunnels in strong earthquake zones are affected by geological activity, and the lining structure is prone to hidden deformation, making it difficult for traditional methods to quickly detect internal cavities after an earthquake, and unable to quantify the real-time impact of vibration on damage.

[0003] To address the problems of low efficiency, insufficient accuracy, and inadequate multi-source data fusion in existing technologies for detecting cavities in tunnel linings in strong earthquake zones, this invention provides a three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones, achieving high-precision and high-efficiency detection. Summary of the Invention

[0004] This invention provides a three-dimensional lidar scanning system for tunnel lining cavities in earthquake-prone areas. This system enables high-precision and high-efficiency scanning and detection of cavities in tunnel linings within earthquake-prone zones, improving the accuracy and reliability of detection. Through the collaborative work of an autonomous navigation robot, measuring instruments, and the three-dimensional lidar, it can rapidly acquire tunnel contour point cloud data. Furthermore, by combining this data with collaborative positioning data obtained from rotational speed sensors and laser displacement sensors, the accuracy and completeness of the data are further improved. The application of a data fusion module enables the organic fusion of multi-period data, providing a solid foundation for establishing a standard tunnel contour model. The construction of this standard tunnel contour model provides an accurate reference for the detection of lining cavities.

[0005] This invention provides a three-dimensional lidar scanning system for tunnel lining cavities in areas of severe earthquakes, comprising:

[0006] The tunnel contour point cloud data acquisition module is used to collect and acquire tunnel contour point cloud data based on the configured autonomous navigation robot, measuring instruments and 3D LiDAR;

[0007] The collaborative positioning data acquisition module is used to acquire collaborative positioning data based on a rotation speed sensor and a laser displacement sensor.

[0008] The data fusion module is used to fuse multi-period tunnel contour point cloud data and cooperative positioning data based on a relative positioning algorithm to obtain fused data.

[0009] The standard tunnel profile model building module is used to process fused data using radial basis neural networks and clustering algorithms to establish a standard tunnel profile model.

[0010] The lining void detection module is used to identify and detect lining voids based on a standard tunnel profile model and measured tunnel profile point cloud data.

[0011] Furthermore, the tunnel contour point cloud data acquisition module includes a data acquisition equipment configuration unit and a data acquisition implementation unit;

[0012] The data acquisition equipment configuration unit is used to configure the autonomous navigation robot, measuring instruments, and 3D LiDAR. The autonomous navigation robot is configured to use the SLAM algorithm for trackless navigation in the tunnel and to achieve localization through fusion of the 3D LiDAR and IMU. The measuring instruments include an acoustic flaw detector and an infrared thermal imager. The 3D LiDAR is configured to perform scanning in a pulse scanning mode according to the set scanning accuracy and maximum scanning frequency.

[0013] The data acquisition implementation unit is used to acquire tunnel contour point cloud data based on the configured autonomous navigation robot, measuring instruments and 3D LiDAR.

[0014] Furthermore, based on the configured autonomous navigation robot, measuring instruments, and 3D LiDAR, tunnel contour point cloud data is acquired, including:

[0015] The autonomous navigation robot is controlled to move along a preset path inside the tunnel, while a 3D LiDAR is activated to scan and obtain 3D LiDAR point cloud data.

[0016] Data on internal cracks in tunnel lining were collected using an acoustic flaw detector.

[0017] Infrared thermal imagers were used to collect data on the seepage areas in the tunnel.

[0018] By combining 3D lidar point cloud data, internal crack data of tunnel lining, and data of water seepage areas in the tunnel, the tunnel outline point cloud data is obtained.

[0019] Furthermore, controlling the autonomous navigation robot to move along a preset path within the tunnel includes: constructing the preset path based on an improved A* algorithm, the cost function of which is defined as:

[0020] ,

[0021] This represents the actual path cost from the starting point to node n. This represents the heuristically estimated cost from the current node n to the destination. The deformation risk coefficient representing node n; This represents the energy consumption cost at node n; , These represent the weighting coefficients for deformation risk and energy consumption cost, respectively, and are automatically adjusted based on task priority.

[0022] Among them, the deformation risk coefficient The risk contribution coefficient is obtained by weighted averaging of the risk contributions of all deformation points within the tunnel. Risk contribution is defined as the product of the reciprocal of the distance between a deformation point and the current node n, and the risk level of that deformation point. Deformation points that are closer to each other and have higher risk levels contribute more to the risk of the current node n. The specific calculation formula is as follows:

[0023] ,

[0024] in, This represents the total number of deformation points detected inside the tunnel. The deformation at the i-th deformation point, Represents the safety threshold of deformation. Deformation point The degree of deformation relative to the safety threshold of deformation; Deformation point The risk to node n decreases with distance; Represents the deformation point; Represents node n and deformation point Euclidean distance This represents the risk attenuation radius, used to control the scope of risk impact.

[0025] Furthermore, based on the rotation speed sensor and the laser displacement sensor, collaborative positioning data is acquired, including:

[0026] Based on the rotation speed sensor, the kilometer marker data of the tunnel is obtained;

[0027] Based on laser displacement sensors, track characteristic data of the tunnel is obtained;

[0028] Based on kilometer marker data and track feature data, collaborative positioning data is generated.

[0029] Furthermore, based on a relative positioning algorithm, multi-period tunnel contour point cloud data and cooperative positioning data are fused to obtain fused data, including:

[0030] The formula for correcting mileage errors using kilometer marker data is as follows:

[0031] ,

[0032] In the above formula, Represents the error compensation coefficient. This represents the corrected mileage error. This represents the preset rotational speed of the equipment under ideal conditions. This represents the actual rotational speed measured in real time by the speed sensor. Represents the original mileage. Represents the baseline mileage;

[0033] Optimize coordinate positioning using trajectory feature data from a laser displacement sensor;

[0034] ,

[0035] in, Let be the measurement deviation correction coefficient of the laser displacement sensor at time j, representing the contribution weight of the j-th laser ranging point to the final calibration; Let be the distance to the tunnel wall measured by the laser displacement sensor at the j-th sampling point. The pre-acquired standard geometric parameters of the track serve as the calibration reference; This represents the coordinates after coarse calibration. Coarse calibration is used to resolve large-scale mileage accumulation errors and relies on the speed sensor and the reference mileage. This represents the coordinates after fine calibration, which is used to eliminate local geometric deviations and relies on laser displacement sensors and orbital feature data.

[0036] Furthermore, the standard tunnel outline model construction module includes a tunnel ordinary inner wall sub-model construction unit, a tunnel pipeline area sub-model construction unit, and a model summary unit;

[0037] The tunnel ordinary inner wall sub-model building unit is used to perform multiple training on fused data based on radial basis function neural network, and remove large error data to establish a tunnel ordinary inner wall sub-model;

[0038] The tunnel pipeline area sub-model building unit is used to train the discarded large error data based on the DBSCAN clustering algorithm to establish the tunnel pipeline area sub-model.

[0039] The model aggregation unit is used to aggregate the ordinary inner wall sub-model of the tunnel and the pipeline area sub-model of the tunnel to obtain the standard tunnel outline model.

[0040] Furthermore, multiple training processes are performed on the fused data based on radial basis function neural networks, including:

[0041] The fused data is normalized and divided into training and testing datasets;

[0042] The training dataset is input into the set radial basis neural network, and the output obtains the standard contour parameters of the tunnel inner wall;

[0043] During training, the prediction residual for each data point is calculated in real time, and data with prediction residuals greater than the set dynamic residual threshold are dynamically removed.

[0044] Furthermore, the lining void detection module includes a detection report generation unit, a detection data storage unit, and a detection result display unit;

[0045] The inspection report generation unit is used to input the acquired measured tunnel contour point cloud data into a standard tunnel contour model, obtain the lining void detection results, and generate an inspection report; the lining void detection results include macro void detection results and micro crack detection results;

[0046] The detection data storage unit is used to store the lining void detection process data and lining void detection results based on distributed encrypted storage technology, so as to realize distributed storage and access control.

[0047] The detection result display unit is used to generate a three-dimensional heat map of the lining void distribution based on the detection results of the lining voids, and to display the actual tunnel scene and the location of the lining voids.

[0048] Furthermore, it also includes a lining void deformation tracking and monitoring module, which is used to track and monitor the deformation trend of the lining void based on the lining void detection results and combined with external vibration monitoring data. Based on the tracking and monitoring results, it generates the deformation risk level of the lining void and the corresponding response strategy. The lining void deformation tracking and monitoring module includes a monitoring data acquisition unit, a deformation risk level determination unit, and a response strategy formulation unit.

[0049] The monitoring data acquisition unit is used to collect deformation monitoring data of micro-cracks based on fiber optic grating sensor arrays and ultrasonic flaw detectors; to collect deformation monitoring data of macro-cavities based on three-dimensional lidar and infrared thermal imagers; and to collect external vibration data of the tunnel area based on triaxial accelerometers and ground acoustic sensors. The unit summarizes the deformation monitoring data of micro-cracks, deformation monitoring data of macro-cavities, and external vibration data of the tunnel area to obtain monitoring data.

[0050] The deformation risk level determination unit is used to analyze monitoring data based on a hybrid deep learning model and output the deformation risk level of the tunnel. Specifically, it uses the convolutional neural network branch in the hybrid deep learning model to analyze the deformation monitoring data of macroscopic cavities to obtain the probability of macroscopic cavity collapse and the impact range on tunnel operation; the input data dimension of the convolutional neural network branch is 200x200x3, and the output data dimension is the collapse probability scalar. It uses the long short-term memory network branch in the hybrid deep learning model to analyze the deformation monitoring data of microscopic cracks and the external vibration data of the tunnel area to predict the microscopic crack propagation trend in the future set period; the input data dimension of the long short-term memory network branch is 120 time steps × 6 sensor channels, and the output dimension is the crack propagation speed. Using the fusion layer in the hybrid deep learning model, through dynamic weighting via an attention mechanism, the deformation risk level of the tunnel is obtained based on the probability of macroscopic cavity collapse, the impact range on tunnel operation, and the microscopic crack propagation trend in the future set period.

[0051] The response strategy formulation unit is used to calculate the stress distribution of the tunnel under different vibration frequencies using the finite element method and construct a vibration impact database; combined with external vibration data of the tunnel area, it assesses and obtains tunnel structural damage risk data based on the vibration impact database; based on the vibration impact database and tunnel structural damage risk data, it weights the deformation risk level to obtain an optimized deformation risk level; and constructs a matching response strategy based on the optimized deformation risk level.

[0052] Compared with existing technologies, this invention has the following advantages and beneficial effects: it can achieve high-precision and high-efficiency scanning and detection of tunnel lining cavities in strong earthquake zones, improving the accuracy and reliability of detection; through the collaborative work of autonomous navigation robots, measuring instruments, and three-dimensional lidar, it can quickly collect tunnel contour point cloud data, and further improve the accuracy and completeness of data by combining the collaborative positioning data obtained by rotation speed sensors and laser displacement sensors; the application of the data fusion module realizes the organic fusion of multi-period data, providing a solid foundation for establishing a standard tunnel contour model; and the construction of the standard tunnel contour model provides an accurate reference for the detection of lining cavities; compared with traditional methods, the cavity positioning error is ≤0.5m, the detection efficiency is improved by 3 times, and the accuracy rate of post-earthquake deformation risk prediction reaches over 92%.

[0053] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0054] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0055] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0056] Figure 1 This is a schematic diagram of a three-dimensional lidar tunnel lining cavity scanning system in a strong earthquake zone, showing the connection relationship between the tunnel contour point cloud data acquisition module, the collaborative positioning data acquisition module, the data fusion module, the standard tunnel contour model construction module, and the lining cavity detection module.

[0057] Figure 2 This is a schematic diagram of the tunnel contour point cloud data acquisition module, showing the composition of the acquisition equipment configuration unit and the data acquisition implementation unit;

[0058] Figure 3 This is a schematic diagram of the standard tunnel outline model construction module, showing the relationship between the ordinary inner wall sub-model construction unit, the pipeline area sub-model construction unit, and the model summary unit. Detailed Implementation

[0059] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0060] This invention provides a three-dimensional lidar scanning system for tunnel lining cavities in areas of severe earthquakes, such as... Figure 1 As shown, it includes:

[0061] The tunnel contour point cloud data acquisition module is used to collect and acquire tunnel contour point cloud data based on the configured autonomous navigation robot, measuring instruments and 3D LiDAR;

[0062] The collaborative positioning data acquisition module is used to acquire collaborative positioning data based on a rotation speed sensor and a laser displacement sensor.

[0063] The data fusion module is used to fuse multi-period tunnel contour point cloud data and cooperative positioning data based on a relative positioning algorithm to obtain fused data.

[0064] The standard tunnel profile model building module is used to process fused data using radial basis neural networks and clustering algorithms to establish a standard tunnel profile model.

[0065] The lining void detection module is used to identify and detect lining voids based on a standard tunnel profile model and measured tunnel profile point cloud data.

[0066] The working principle of the above technical solution is as follows: In order to realize a three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones, the various modules of this invention work closely together, and their working principle is as follows:

[0067] First, the tunnel contour point cloud data acquisition module accomplishes this task based on a configured autonomous navigation robot, measuring instruments, and a 3D LiDAR. The autonomous navigation robot can autonomously plan its path in the complex tunnel environment. It is equipped with measuring instruments and a 3D LiDAR and travels along the tunnel. The 3D LiDAR quickly acquires the spatial coordinate information of various points on the tunnel surface by emitting laser beams and receiving the reflected signals, thus forming point cloud data. The measuring instruments assist the LiDAR to improve the accuracy of data acquisition. For example, in a mountain tunnel located in a strong earthquake zone, the autonomous navigation robot starts from the tunnel entrance and slowly moves forward according to a preset route. The 3D LiDAR continuously emits laser beams to scan various parts of the tunnel, such as the top and side walls. Every reflected laser signal is recorded, ultimately forming the complete contour point cloud data of the tunnel, providing a foundation for subsequent analysis.

[0068] Secondly, to more accurately determine the scanning position and orientation, the collaborative localization data acquisition module uses a rotation speed sensor and a laser displacement sensor to collect collaborative localization data. The rotation speed sensor measures the wheel speed of the autonomous navigation robot, and by calculating the rotation speed and time, the robot's moving distance and direction can be deduced. The laser displacement sensor measures the distance between the robot and the tunnel wall to further determine the robot's position. Taking the mountain tunnel mentioned earlier as an example, when the autonomous navigation robot is traveling in the tunnel, the rotation speed sensor monitors the wheel rotation in real time. If the wheel speed increases within a certain distance, it indicates that the robot's travel speed is increasing, and vice versa. The laser displacement sensor continuously measures the distance between the robot and the tunnel sidewall. When the distance changes, it indicates that the robot's position or orientation has changed. This collaborative localization data, together with the tunnel contour point cloud data, provides important information for subsequent data fusion.

[0069] The collected tunnel contour point cloud data and collaborative localization data are independent. To better analyze and process them, they need to be fused. The data fusion module, based on a relative positioning algorithm, processes multi-period tunnel contour point cloud data and collaborative localization data. Through the relative positioning algorithm, data collected at different times and locations can be matched and aligned, eliminating errors and inconsistencies between data, thus obtaining more accurate and complete fused data. Assuming that during the scanning process of the aforementioned mountain tunnel, due to the complex tunnel environment, the autonomous navigation robot may experience slight shaking or deviation during its movement, leading to certain errors in the collected point cloud data and collaborative localization data, the data fusion module uses the relative positioning algorithm to compare and analyze the multi-period data, identify the relative relationships between the data, and fuse them into a unified dataset. This provides more accurate tunnel information and a reliable data foundation for subsequent model building.

[0070] The standard tunnel profile model construction module utilizes radial basis function neural networks (RBNs) and clustering algorithms to process fused data and establish a standard tunnel profile model. RBNs are powerful machine learning algorithms capable of nonlinearly fitting complex data and identifying inherent patterns. Clustering algorithms classify fused data based on similarity, grouping data points with the same characteristics into one category. In the example of a mountain tunnel, through analysis of the fused data, RBNs can learn features such as the shape and size of the tunnel profile. The clustering algorithm classifies the points in the fused data, for example, grouping data points from different parts of the tunnel, such as the top and sidewalls, into different categories. Then, based on these classification results and the learned features, a standard profile model of the tunnel is established. This model can serve as a reference standard for subsequent lining void detection.

[0071] The lining void detection module is based on a standard tunnel contour model. It identifies and detects lining voids using measured tunnel contour point cloud data. During actual inspection, the measured tunnel contour point cloud data is compared with the standard tunnel contour model. If there is a significant difference between the measured data and the standard model, it indicates that a lining void may exist in that area. Taking a mountain tunnel as an example, after the autonomous navigation robot completes a scan, it compares the collected measured tunnel contour point cloud data with the previously established standard tunnel contour model. If, in a certain section of the tunnel, the measured data shows a significant deviation between the contour of that area and the standard model, such as the appearance of local depressions or voids, then it can be determined that a lining void exists in that area. Based on the detection results, staff can promptly repair the lining void to ensure tunnel safety.

[0072] The beneficial effects of the above technical solution are as follows: By adopting the solution provided in this embodiment, through a highly integrated and automated data acquisition, processing and analysis process, rapid and accurate detection of tunnel lining voids is achieved, providing strong support for the safe operation and maintenance of tunnels.

[0073] In one embodiment, such as Figure 2 As shown, the tunnel contour point cloud data acquisition module includes a data acquisition device configuration unit and a data acquisition implementation unit;

[0074] The data acquisition configuration unit is used to configure the autonomous navigation robot, measuring instruments, and 3D LiDAR. The autonomous navigation robot is configured to use the SLAM algorithm for trackless navigation within the tunnel and to achieve localization through fusion positioning of the 3D LiDAR and IMU. The measuring instruments include an acoustic flaw detector and an infrared thermal imager. The 3D LiDAR is configured to perform pulse scanning according to the set scanning accuracy and maximum scanning frequency. When configuring the 3D LiDAR, the specific scanning accuracy can typically be set to the millimeter level, such as 5 mm or 10 mm. The maximum scanning frequency can be set between 10,000 Hz and 20,000 Hz, enabling the acquisition of a sufficient number of data points per unit time to comprehensively and quickly complete the scanning of the tunnel walls. When the autonomous navigation robot uses the SLAM algorithm for trackless navigation, its positioning accuracy can be controlled at the centimeter level, for example, a horizontal positioning accuracy of ±5 cm and a vertical positioning accuracy of ±3 cm, to ensure that the robot can move stably and accurately within the tunnel along a predetermined path.

[0075] The data acquisition implementation unit is used to acquire tunnel contour point cloud data based on the configured autonomous navigation robot, measuring instruments and 3D LiDAR.

[0076] The working principle of the above technical solution is as follows: To realize the function of the tunnel contour point cloud data acquisition module, the autonomous navigation robot of this invention travels in the tunnel according to a preset path, uses the SLAM algorithm to achieve trackless navigation, and combines IMU for fusion positioning to ensure the robot's accurate positioning in the tunnel; the acoustic flaw detector is used to detect material defects and damage in the tunnel lining. By emitting sound waves and receiving the reflected signals, it analyzes the propagation time and amplitude changes of the sound waves to determine the integrity and degree of damage of the lining material; the infrared thermal imager is used to detect the temperature distribution on the surface of the tunnel lining. By capturing infrared radiation and converting it into a visible thermal image, it can intuitively display the temperature differences on the lining surface, which helps to discover potential cavities or heat loss areas; the three-dimensional lidar uses a pulse scanning method to quickly scan the inner wall of the tunnel according to the set scanning accuracy and maximum scanning frequency to acquire high-precision point cloud data. This data can reflect the shape and size of the tunnel contour in detail; the data acquisition implementation unit is responsible for coordinating the work of each acquisition device to ensure synchronous acquisition and complete recording of data, providing a reliable foundation for subsequent data processing and analysis.

[0077] The beneficial effects of the above technical solution are as follows: The solution provided in this embodiment enables comprehensive, efficient, and accurate scanning and detection of cavities in the tunnel lining during strong earthquakes; the precise positioning and navigation of the autonomous navigation robot ensures the stability and reliability of the scanning system, allowing it to operate continuously in complex tunnel environments; the introduction of acoustic flaw detectors and infrared thermal imagers provides the system with multiple detection methods, enabling evaluation of the tunnel lining from multiple angles and improving the accuracy and comprehensiveness of the detection; the high-precision scanning of the three-dimensional lidar can acquire detailed tunnel contour point cloud data, providing rich information for subsequent data processing and analysis; the coordinating role of the data acquisition implementation unit ensures efficient collaboration between various acquisition devices, improving the overall efficiency and performance of the system.

[0078] In one embodiment, based on a configured autonomous navigation robot, measuring instruments, and a 3D LiDAR, tunnel contour point cloud data is acquired, including:

[0079] The autonomous navigation robot is controlled to move along a preset path inside the tunnel, while a 3D LiDAR is activated to scan and obtain 3D LiDAR point cloud data.

[0080] Data on internal cracks in tunnel lining are collected using an acoustic flaw detector. When the acoustic flaw detector is working, its emission frequency can be set in the range of 2 MHz to 5 MHz. This frequency range can penetrate the tunnel lining structure well and detect any small defects that may exist inside.

[0081] Infrared thermal imagers are used to collect data on seepage areas in tunnels. The temperature resolution of infrared thermal imagers can reach 0.05℃, which can keenly capture minute temperature changes on the lining surface and promptly detect potential cavities.

[0082] By combining 3D lidar point cloud data, internal crack data of tunnel lining, and data of water seepage areas in the tunnel, the tunnel outline point cloud data is obtained.

[0083] The working principle of the above technical solution is as follows: First, the invention controls an autonomous navigation robot to move precisely inside the tunnel according to a pre-set path and navigation information. During this movement, a 3D lidar begins operation, rapidly capturing the 3D coordinate information of the tunnel wall using a pulsed scanning method, forming high-precision point cloud data. This data records in detail the shape and size of the tunnel outline, as well as any possible deformations or anomalies. Simultaneously, an acoustic flaw detector analyzes the cracks inside the tunnel lining by emitting sound waves and receiving the reflected signals. Sound waves encountering cracks during propagation will be reflected and scattered. By analyzing these signals, the flaw detector can accurately determine the location and size of the cracks. The infrared thermal imager uses the principle of infrared radiation to detect the temperature distribution inside the tunnel. In areas of seepage, the evaporation of water absorbs heat from the surrounding environment, causing the temperature to drop and creating a clear temperature difference area on the infrared thermal image. The infrared thermal imager can capture this temperature difference information, providing a strong basis for judging the tunnel seepage situation. The three-dimensional lidar point cloud data provides a precise description of the tunnel outline, the acoustic flaw detector data reveals the cracks inside the tunnel lining, and the infrared thermal imager data reflects the location and range of the seepage area in the tunnel. These data together constitute the complete tunnel outline point cloud data, providing a comprehensive and accurate information foundation for subsequent data processing and analysis.

[0084] The beneficial effects of the above technical solution are as follows: the solution provided in this embodiment can achieve comprehensive, efficient, and non-contact scanning of tunnel lining cavities in strong earthquake zones, greatly improving the accuracy and efficiency of the scanning; the use of autonomous navigation robots ensures the automation and intelligence of the scanning process, reduces manual intervention, and improves operational safety; the high-precision scanning capability of three-dimensional lidar can capture minute changes in the tunnel inner wall, providing the possibility for timely detection and handling of potential safety hazards; the introduction of acoustic flaw detectors and infrared thermal imagers further enriches the dimensions of the scanning data, allowing for a direct presentation of internal cracks and water seepage in the tunnel lining, providing strong support for the comprehensive assessment of the tunnel structure.

[0085] In one embodiment, controlling an autonomous navigation robot to move along a preset path within a tunnel includes: constructing the preset path based on an improved A* algorithm, wherein the cost function of the improved A* algorithm is defined as:

[0086] ,

[0087] This represents the actual path cost from the starting point to node n. This represents the heuristically estimated cost from the current node n to the destination. The deformation risk coefficient representing node n; This represents the energy consumption cost at node n; , These represent the weighting coefficients for deformation risk and energy consumption cost, respectively, and are automatically adjusted based on task priority.

[0088] Among them, the deformation risk coefficient The risk contribution coefficient is obtained by weighted averaging of the risk contributions of all deformation points within the tunnel. Risk contribution is defined as the product of the reciprocal of the distance between a deformation point and the current node n, and the risk level of that deformation point. Deformation points that are closer to each other and have higher risk levels contribute more to the risk of the current node n. The specific calculation formula is as follows:

[0089] ,

[0090] in, This represents the total number of deformation points detected inside the tunnel. The deformation at the i-th deformation point, The representative deformation safety threshold is set according to the "Railway Tunnel Lining Quality Inspection Specification" (TB 10223-2022). Deformation point The degree of deformation relative to the safety threshold of deformation; Deformation point The risk to node n decreases with distance; Represents the deformation point; Represents node n and deformation point Euclidean distance Represents the risk attenuation radius, used to control the range of risk impact, and is determined by fitting tunnel geological exploration data.

[0091] The working principle of the above technical solution is as follows: The core of the improved A* algorithm in constructing the preset path is to comprehensively consider various factors through a cost function in order to plan an optimal path suitable for the tunnel environment.

[0092] The actual path cost represents the actual path cost from the starting point to node n. In a tunnel environment, the actual path cost may involve factors such as distance and terrain complexity. For example, in a relatively straight tunnel section, the actual path cost may mainly depend on the distance traveled. However, in areas with curves and changes in slope, in addition to distance, factors such as the difficulty of turning and the ease of climbing or descending slopes must also be considered, all of which will increase the actual path cost.

[0093] Heuristic cost estimation is the heuristic estimated cost from the current node n to the destination. It is usually estimated using methods such as Euclidean distance. It provides the algorithm with a general direction guide, leading the autonomous navigation robot toward the destination. In tunnel scenarios, it allows the algorithm to quickly determine the general direction of travel and avoids blind exploration during the search process.

[0094] The deformation risk coefficient R(n) is obtained by weighted averaging of the risk contributions of all deformation points within the tunnel. The risk contribution of each deformation point is the product of the reciprocal of the distance between the deformation point and the current node n and the risk level of the deformation point. The closer the deformation point is and the higher its risk level, the greater its risk contribution to the current node n. This means that when planning the path, priority will be given to avoiding deformation areas that are close to the point and have a high risk level. At the same time, it is calculated accurately using a specific formula.

[0095] Considering the energy consumption of autonomous navigation robots at different nodes, such as when climbing or turning, energy consumption will increase. In path planning, energy cost is used to reflect these differences in energy consumption, so that the planned path can minimize energy consumption.

[0096] The weighting coefficients of deformation risk coefficient and energy consumption cost are automatically adjusted according to task priority. If the task focuses more on safety, the weighting coefficient of deformation risk coefficient will be increased, making the path planning more inclined to avoid high-risk areas; if the energy consumption requirement is high, the weighting coefficient of energy consumption cost will be increased, making the path more energy-efficient.

[0097] The beneficial effects of the above technical solution are as follows: The solution provided in this embodiment enables precise scanning and deformation risk assessment of tunnel lining cavities in areas prone to strong earthquakes; the autonomous navigation robot's mobile scanning within the tunnel, combined with the optimal path planned using the improved A* algorithm, not only improves scanning efficiency but also ensures the safety of the scanning process; simultaneously, by collecting real-time deformation data of the tunnel lining and calculating the deformation risk coefficient, the risk contribution of each deformation point within the tunnel to the current node can be accurately assessed, providing strong data support for tunnel maintenance and repair; furthermore, this invention considers factors such as the deformation amount of the deformation point, the deformation safety threshold, and the distance between the deformation point and the current node, and by introducing a risk attenuation radius, further improves the accuracy and reliability of the assessment results, providing strong protection for the safe operation of tunnel engineering.

[0098] In one embodiment, collaborative positioning data is acquired based on a rotation speed sensor and a laser displacement sensor, including:

[0099] Based on the rotation speed sensor, the kilometer marker data of the tunnel is obtained;

[0100] Based on laser displacement sensors, track characteristic data of the tunnel is obtained;

[0101] Based on kilometer marker data and track feature data, collaborative positioning data is generated.

[0102] The working principle of the above technical solution is as follows: The rotation speed sensor monitors the rotation speed of the equipment inside the tunnel and, combined with a preset mileage calculation model, can accurately calculate the kilometer marker data at any location inside the tunnel. The laser displacement sensor emits a laser beam and receives the reflected signal. Based on the round-trip time and angle change of the laser beam, it accurately measures the geometric features of the tunnel track, such as the elevation and direction of the track. These two types of data complement each other and together form the basis of collaborative positioning data. In the process of summarizing and generating collaborative positioning data, this invention preprocesses the raw data collected by the rotation speed sensor and the laser displacement sensor, including noise reduction, filtering, and calibration steps, to ensure the accuracy and consistency of the data. Subsequently, according to the preset algorithm model, the kilometer marker data and track feature data are fused to generate collaborative positioning data containing precise location information and track geometric features, providing reliable data support for subsequent 3D scanning and deformation risk assessment.

[0103] The beneficial effects of the above technical solution are as follows: the solution provided in this embodiment can achieve high-precision scanning and positioning of cavities in the lining of tunnels in strong earthquake zones; through the collaborative work of rotation speed sensors and laser displacement sensors, kilometer marker data and track geometric features at any location within the tunnel can be acquired in real time, and the high precision and real-time nature of this data provides a solid foundation for subsequent scanning and analysis; preprocessing of the raw data effectively removes noise and interference, improving the accuracy and reliability of the data; through a preset algorithm model, kilometer marker data and track feature data are fused to generate collaborative positioning data containing precise location information and track geometric features.

[0104] In one embodiment, based on a relative positioning algorithm, multi-period tunnel contour point cloud data and cooperative positioning data are fused to obtain fused data, including:

[0105] The formula for correcting mileage errors using kilometer marker data is as follows:

[0106] ,

[0107] In the above formula, Represents the error compensation coefficient. This represents the corrected mileage error. This represents the preset rotational speed of the equipment under ideal conditions. This represents the actual rotational speed measured in real time by the speed sensor. Represents the original mileage. Represents the baseline mileage;

[0108] Optimize coordinate positioning using trajectory feature data from a laser displacement sensor;

[0109] ,

[0110] in, Let be the measurement deviation correction coefficient of the laser displacement sensor at time j, representing the contribution weight of the j-th laser ranging point to the final calibration; Let be the distance to the tunnel wall measured by the laser displacement sensor at the j-th sampling point. The pre-acquired standard geometric parameters of the track serve as the calibration reference; This represents the coordinates after coarse calibration. Coarse calibration is used to resolve large-scale mileage accumulation errors and relies on the speed sensor and the reference mileage. This represents the coordinates after fine calibration, which is used to eliminate local geometric deviations and relies on laser displacement sensors and orbital feature data.

[0111] The working principle of the above technical solution is as follows: In tunnels located in areas prone to strong earthquakes, the detection of lining cavities faces numerous challenges due to complex geological conditions. During tunnel scanning, the actual operating conditions of the equipment may differ from the ideal state. Ideally, the equipment would operate at a preset speed, thus generating a theoretical mileage. However, in reality, the speed may be affected by various factors, such as tunnel surface conditions and equipment mechanical wear, leading to a deviation between the actual and preset speeds. This speed deviation will further cause errors in mileage calculation. Kilometer marker data are fixed and accurate mileage markers in the tunnel. By using the error compensation coefficient β, combined with the ideal speed, actual speed, original mileage, and reference mileage, the original mileage can be corrected. The error compensation coefficient β can be determined based on a large amount of experimental data or experience in practical applications. It reflects the degree of compensation for the mileage error caused by the speed deviation.

[0112] Scanning tunnel lining cavities requires precise coordinate positioning. When processing large-scale mileage accumulation errors, a coarse calibration method is used, relying on speed sensors and reference mileage. However, this coarse calibration cannot handle local geometric deviations because the actual conditions of tunnels are complex, and the local shape of the track may differ from the standard geometric parameters. Laser displacement sensors can measure the distance to the tunnel wall in real time. The pre-collected standard track geometric parameters are used as the calibration benchmark. The distance to the tunnel wall measured by the laser displacement sensor at each sampling point is compared with the standard track geometric parameters. By measuring the deviation correction coefficient, the contribution weight of each sampling point to the final calibration is determined. Thus, fine calibration is performed on the basis of coarse calibration coordinates, eliminating local geometric deviations and making the coordinate positioning more accurate.

[0113] After obtaining the corrected and optimized collaborative positioning data (i.e., the corrected mileage error and finely calibrated coordinates) through the above steps, the multi-cycle tunnel contour point cloud data collected by the three-dimensional lidar can be associated with the corresponding precise spatiotemporal location to form the fused data.

[0114] In a specific application, assuming a tunnel scanning project, the baseline mileage of a scanning device is 1000 meters, the original mileage is 500 meters, the preset rotation speed of the device under ideal conditions is 100 rpm, the actual rotation speed measured by the rotation speed sensor in real time is 90 rpm, and the error compensation coefficient β is 0.8; according to the calculation formula, the corrected mileage error is 580 meters. This shows that because the actual rotation speed is lower than the ideal rotation speed, there is an error in the original mileage. After correction, more accurate mileage data is obtained.

[0115] The beneficial effects of the above technical solution are as follows: the solution provided in this embodiment not only improves measurement efficiency, but also significantly enhances the accuracy and reliability of measurement, providing strong technical support for the detection of lining voids in tunnels in strong earthquake zones.

[0116] In one embodiment, such as Figure 3 As shown, the standard tunnel outline model construction module includes a tunnel ordinary inner wall sub-model construction unit, a tunnel pipeline area sub-model construction unit, and a model summary unit;

[0117] The tunnel ordinary inner wall sub-model building unit is used to perform multiple training on fused data based on radial basis function neural network and remove large error data to establish a tunnel ordinary inner wall sub-model. When the radial basis function neural network processes fused data to establish a standard tunnel contour model, the number of hidden layer nodes can be set from 10 to 30 according to the specific tunnel conditions and data complexity, and the learning rate can be set between 0.01 and 0.1. By continuously adjusting these parameters, the model can more accurately fit the tunnel contour features.

[0118] The tunnel pipeline area sub-model construction unit is used to train the discarded large error data based on the DBSCAN clustering algorithm to establish the tunnel pipeline area sub-model. When processing data, the number of cluster centers can be set to 5 to 15 according to the tunnel segmentation and geological characteristics, and the cluster radius can be set to the range of 10 cm to 50 cm according to the data distribution to achieve effective classification and processing of data, and provide support for establishing an accurate standard tunnel outline model.

[0119] The model aggregation unit is used to aggregate the ordinary inner wall sub-model of the tunnel and the pipeline area sub-model of the tunnel to obtain the standard tunnel outline model.

[0120] The working principle of the above technical solution is as follows: In the process of constructing a standard tunnel contour model, this invention first constructs units based on the ordinary inner wall sub-model of the tunnel, and uses a radial basis function neural network (RBN) to perform multiple trainings on the fused data. The RBN has a powerful nonlinear mapping capability and can handle complex data relationships. When processing the fused data, the number of hidden layer nodes can be set from 10 to 30 depending on the specific tunnel conditions and data complexity. For example, for a short tunnel with relatively simple geological conditions, its data complexity is low, and the number of hidden layer nodes can be set to 10. This ensures the model's ability to fit the data while avoiding excessive computation. However, for a long tunnel with complex geological conditions, the data complexity is high, and the number of hidden layer nodes can be set to... To better capture the complex features of the tunnel inner wall, 30 learning rates are used. The learning rate can be set between 0.01 and 0.1. The learning rate controls the step size of parameter updates during model training. If the learning rate is set too high, the model may skip the optimal solution during training, resulting in inaccurate training results. If the learning rate is set too low, the model training speed will become very slow. By continuously adjusting these parameters, the model can more accurately fit the tunnel contour features. During training, the model will continuously compare the predicted results with the actual data and adjust its own parameters according to the error. After multiple training sessions, some data points with large errors may be caused by random errors of the sensors or special interference in the tunnel. These large error data will be removed, thereby establishing an accurate sub-model of the ordinary inner wall of the tunnel.

[0121] Taking a subway tunnel in a certain city as an example, during the construction process, in order to build a standard tunnel outline model, LiDAR and displacement sensors were used to collect data on the inner wall of the tunnel. Since the subway tunnel is of moderate length but the geological conditions are somewhat complex, with some alternating rock layers and soft soil layers, when building the ordinary inner wall sub-model of the tunnel, the technicians set the number of hidden layer nodes of the radial basis neural network to 20 and the learning rate to 0.05. After multiple training and parameter adjustments, the model can fit the outline features of the inner wall of the tunnel very well. During the training process, it was found that some data points deviated significantly from the model's prediction results. After analysis, these large error data were found to be due to the random errors caused by the LiDAR passing through some construction-remaining obstacles in the tunnel. After removing these large error data, the established ordinary inner wall sub-model of the tunnel can accurately reflect the real situation of the ordinary inner wall of the subway tunnel.

[0122] The next step is the sub-model construction unit for the tunnel pipeline area. This unit trains on the discarded large-error data using the DBSCAN clustering algorithm. This large-error data often contains information about the tunnel pipeline area because the structure and characteristics of pipelines (such as cables and water pipes) inside the tunnel differ from those of ordinary tunnel walls, leading to significant deviations in the data collected by sensors. The DBSCAN clustering algorithm is a density-based spatial clustering algorithm that divides data into different clusters based on the density of data points. When processing the data, the number of cluster centers can be set from 5 to 15, depending on the tunnel's segmentation and geological characteristics. For data points with large prediction residuals, DBSCAN clustering analysis can determine whether they form clusters with certain characteristics. If such clusters exist, further analysis of the specific conditions of the tunnel lining represented by these clusters is possible. For example, these clusters may correspond to concentrated areas of tunnel lining voids or other abnormal structures in the tunnel lining. Detailed study of the geometric characteristics and spatial distribution of the clusters can yield more precise information about the location, size, and shape of tunnel lining voids. Furthermore, the DBSCAN clustering algorithm can also be used to... By summarizing the distribution patterns of these outlier data points and considering factors such as the tunnel's geological environment and seismic zone characteristics, a more comprehensive correlation model between tunnel lining voids and geological conditions can be established. For example, for a tunnel divided into multiple functional areas (such as driving areas and equipment areas), the geological characteristics and pipeline distribution of each area may differ. If the geological conditions of a certain area are relatively stable and the pipeline distribution is relatively concentrated, fewer cluster centers can be set, such as 5. Conversely, for areas with complex geological conditions and dispersed pipeline distribution, more cluster centers can be set, such as 15. The cluster radius can be determined based on... The data distribution is set within the range of 10 cm to 50 cm. The cluster radius determines how many other data points are needed around a data point to form a cluster. If the cluster radius is set too small, the data may be divided into too many small clusters, failing to accurately reflect the overall characteristics of the pipeline area. If the cluster radius is set too large, some data points that do not belong to the same pipeline area may be merged together, causing inaccuracy in the model. By reasonably setting the number of cluster centers and the cluster radius, the algorithm can effectively classify and process the large error data that has been removed, thereby establishing a sub-model of the tunnel pipeline area.

[0123] Finally, there is the model aggregation unit, which aggregates the tunnel ordinary inner wall sub-model and the tunnel pipeline area sub-model to obtain the standard tunnel outline model. The standard tunnel outline model integrates the features of the tunnel ordinary inner wall and the pipeline area, and can comprehensively and accurately reflect the actual outline of the tunnel.

[0124] In practical applications, the data obtained by comparing the standard tunnel outline model with the actual scan data is used. For example, during the regular inspection of the subway tunnel, the tunnel is scanned again using equipment such as LiDAR to obtain the actual scan data. This actual scan data is then compared with the standard tunnel outline model. If a deviation is found between the actual data and the model, it may mean that there are problems such as lining voids in the tunnel. Suppose that in a certain inspection, it is found that the actual scan data of a certain section of the tunnel deviates significantly from the standard model at a certain location. After further on-site investigation, it is found that there are indeed lining voids at that location. This shows that by comparing the standard tunnel outline model with the actual scan data, safety hazards in the tunnel can be detected in a timely manner, providing an accurate basis for the maintenance and repair of the tunnel, thereby ensuring the safety and stability of the tunnel in complex environments such as strong earthquake zones.

[0125] The beneficial effects of the above technical solution are as follows: The solution provided in this embodiment can significantly improve the accuracy and efficiency of the tunnel lining cavity scanning system in strong earthquake zones. Through the training of radial basis neural networks, the system can automatically identify and remove large error data, avoiding the influence of complex geological conditions or equipment errors on the scanning results, thereby improving the accuracy of the ordinary inner wall sub-model of the tunnel. At the same time, by using clustering algorithms to conduct in-depth analysis of the removed large error data, the feature points of the tunnel pipeline area can be accurately identified, and corresponding sub-models can be established, further improving the integrity and accuracy of the entire standard tunnel outline model.

[0126] In one embodiment, multiple training of the fused data based on a radial basis function neural network includes:

[0127] The fused data is normalized and divided into training and testing datasets;

[0128] The training dataset is input into the set radial basis neural network, and the output obtains the standard contour parameters of the tunnel inner wall;

[0129] During training, the prediction residual for each data point is calculated in real time, and data with prediction residuals greater than the set dynamic residual threshold are dynamically removed.

[0130] The working principle of the above technical solution is as follows: First, normalization ensures that the fused data is analyzed on a uniform scale, improving the training efficiency and accuracy of the radial basis function neural network. The division of the training dataset and the test dataset is to verify the generalization ability of the model and ensure that the model can work stably and reliably in practical applications. During the training process of the radial basis function neural network, the model gradually approaches the true standard contour parameters of the tunnel inner wall through continuous iterative optimization. At the same time, the prediction residual is calculated in real time and large error data is dynamically removed, which can effectively avoid the interference of these abnormal data on the model training process and improve the robustness of the model.

[0131] The beneficial effects of the above technical solution are as follows: By adopting the solution provided in this embodiment, through multiple training of fused data and dynamic elimination of large error data, the standard contour parameters of the tunnel inner wall are accurately obtained and the potential pipeline area or structural defects are effectively identified.

[0132] In one embodiment, the lining void detection module includes a detection report generation unit, a detection data storage unit, and a detection result display unit;

[0133] The inspection report generation unit is used to input the acquired measured tunnel contour point cloud data into a standard tunnel contour model, obtain the lining void detection results, and generate an inspection report; the lining void detection results include macro void detection results and micro crack detection results;

[0134] The detection data storage unit is used to store the lining void detection process data and lining void detection results based on distributed encrypted storage technology, realizing distributed storage and access control; the blockchain technology specifically adopts a lightweight consortium chain architecture, with each detection node acting as a ledger node, and the crack data hash value is stored on the chain for evidence.

[0135] The detection result display unit is used to generate a three-dimensional heat map of the lining void distribution based on the detection results of the lining voids, and to display the actual tunnel scene and the location of the lining voids.

[0136] The working principle of the above technical solution is as follows: In the workflow of the lining void detection module, the detection report generation unit of this invention can automatically generate a detailed report containing macroscopic void detection results and microscopic crack detection results, providing tunnel maintenance personnel with intuitive and comprehensive detection information; the detection data storage unit utilizes blockchain technology to securely and efficiently store these valuable detection data and the final detection results. Distributed blockchain storage technology, with its decentralized and tamper-proof characteristics, ensures the authenticity and integrity of the data. At the same time, through the permission management mechanism, only authorized personnel can access this sensitive information, effectively protecting data security; the detection result display unit presents the lining void detection results in the form of a three-dimensional distribution heat map. This visualization method not only intuitively shows the distribution of voids, but also indicates the severity of voids through color depth, making it easy for maintenance personnel to quickly identify problem areas; this unit can also overlay the heat map onto the actual tunnel scene, achieving seamless integration of virtual and reality. After wearing AR glasses, maintenance personnel can directly see the location of voids and related information inside the tunnel, greatly improving detection efficiency and accuracy.

[0137] The beneficial effects of the above technical solution are as follows: The solution provided in this embodiment enables efficient and accurate scanning and detection of lining voids in tunnels located in areas of strong earthquakes. It also allows for the overlay display of the actual tunnel scene and the location of the lining voids, enabling maintenance personnel to intuitively understand the distribution and severity of voids without tedious comparisons. This is of great significance for quickly developing repair plans and ensuring the safety of tunnel structures. Furthermore, the data storage solution combined with distributed blockchain storage technology not only ensures the authenticity and integrity of the detection data but also effectively prevents data leakage through access control. It enables mutual trust and sharing of data among multiple organizations (construction / supervision / operation and maintenance parties), prevents tampering with detection reports, and provides strong data support for tunnel management.

[0138] In one embodiment, the system further includes a lining void deformation tracking and monitoring module, which is used to track and monitor the deformation trend of the lining void based on the lining void detection results and combined with external vibration monitoring data, and generate the deformation risk level and response strategy of the lining void according to the tracking and monitoring results; the lining void deformation tracking and monitoring module includes a monitoring data acquisition unit, a deformation risk level determination unit, and a response strategy formulation unit;

[0139] The monitoring data acquisition unit is used to collect deformation monitoring data of micro-cracks based on fiber optic grating sensor arrays and ultrasonic flaw detectors; to collect deformation monitoring data of macro-cavities based on three-dimensional lidar and infrared thermal imagers; and to collect external vibration data of the tunnel area based on triaxial accelerometers and ground acoustic sensors. The unit summarizes the deformation monitoring data of micro-cracks, deformation monitoring data of macro-cavities, and external vibration data of the tunnel area to obtain monitoring data.

[0140] Specifically, the process involves: using a distributed fiber optic grating sensor array on the tunnel wall to monitor and acquire the width, length, and propagation rate of micro-cracks; using an ultrasonic flaw detector to detect and acquire the crack depth and internal micro-cracks; and generating monitoring data for micro-cracks based on the width, length, propagation rate, crack depth, and internal micro-cracks. It also involves using 3D lidar scanning to acquire changes in the volume, location, and morphology of macro-cavities; using an infrared thermal imager to measure temperature field anomalies to identify water seepage or material peeling within macro-cavities; and generating monitoring data for macro-cavities based on these changes. Finally, it involves using a triaxial accelerometer to measure the intensity, frequency, and direction of external vibrations such as earthquakes and traffic vibrations; and using a ground acoustic sensor to capture low-frequency vibration signals caused by geological activity; and generating external vibration data for the tunnel area based on the intensity, frequency, direction, and low-frequency vibration signals.

[0141] The deformation risk level determination unit is used to analyze monitoring data based on a hybrid deep learning model and output the deformation risk level of the tunnel. Specifically, it uses the convolutional neural network branch in the hybrid deep learning model to analyze the deformation monitoring data of macroscopic cavities to obtain the probability of macroscopic cavity collapse and the impact range on tunnel operation; the input data dimension of the convolutional neural network branch is 200x200x3, and the output data dimension is the collapse probability scalar. It uses the long short-term memory network branch in the hybrid deep learning model to analyze the deformation monitoring data of microscopic cracks and the external vibration data of the tunnel area to predict the microscopic crack propagation trend in the future set period; the input data dimension of the long short-term memory network branch is 120 time steps × 6 sensor channels, and the output dimension is the crack propagation speed. Using the fusion layer in the hybrid deep learning model, through dynamic weighting via an attention mechanism, the deformation risk level of the tunnel is obtained based on the probability of macroscopic cavity collapse, the impact range on tunnel operation, and the microscopic crack propagation trend in the future set period.

[0142] The response strategy formulation unit is used to calculate the stress distribution of the tunnel under different vibration frequencies using the finite element method and construct a vibration impact database; combined with external vibration data of the tunnel area, it assesses and obtains tunnel structural damage risk data based on the vibration impact database; based on the vibration impact database and tunnel structural damage risk data, it weights the deformation risk level to obtain an optimized deformation risk level; and constructs a matching response strategy based on the optimized deformation risk level.

[0143] The working principle of the above technical solution is as follows: In the environment of strong earthquake zones, the deformation trend of lining cavities is a key indicator for assessing the structural safety of tunnels. This invention achieves accurate tracking and monitoring of the deformation trend of lining cavities through a lining cavity deformation tracking and monitoring module. First, the monitoring data acquisition unit makes full use of various high-precision sensors, such as fiber optic grating sensor arrays, ultrasonic flaw detectors, three-dimensional lidar, infrared thermal imagers, triaxial accelerometers, and ground acoustic sensors, to comprehensively collect vibration data of micro-cracks, macro-cavities, and the external environment of the tunnel area, ensuring the comprehensiveness and accuracy of the monitoring data. Next, the deformation risk level determination unit uses a hybrid deep learning model to conduct in-depth analysis of the monitoring data. This model combines the advantages of convolutional neural networks and long short-term memory networks, and can accurately identify the collapse probability of macro-cavities, the expansion trend of micro-cracks, and the impact range of tunnel operation, thereby scientifically and rationally outputting the deformation risk level of the tunnel.

[0144] The input data dimension 200x200x3 indicates that the data input to this convolutional neural network branch is a three-dimensional data structure. The first two dimensions, "200x200," represent the spatial dimensions of the image, meaning that the height and width of the image are both 200 pixels. This implies that the input image consists of 200 pixels horizontally and vertically. The third dimension, "3," represents the number of channels in the image. In color images, red (R), green (G), and blue (B) channels are typically used to represent color information, so 3 here indicates that the input is a color image. The collapse probability label... The scalar in the quantity is a numerical value with only magnitude and no direction. This single numerical value represents the "collapse probability". In specific application scenarios, "collapse" may have different meanings. For example, in the field of architecture, it may refer to the collapse of a building, and in the field of geology, it may refer to the collapse of the ground. After the convolutional neural network branch performs a series of convolution, pooling, activation and other operations on the input 200x200x3 image data, it finally outputs a value between 0 and 1, which represents the probability of collapse. For example, an output value of 0.8 means that there is an 80% chance that a collapse will occur.

[0145] The input data dimension of the Long Short-Term Memory (LSTM) network branch consists of 120 time steps and 6 sensor channels. "Time step" represents the number of sampling points in the time series, meaning the data is recorded sequentially, with information from 120 different moments. "Sensor channel" represents the use of 6 different sensors to collect data, with each sensor's data constituting one channel. For example, when monitoring a physical system, 6 different types of sensors (such as accelerometers, displacement sensors, etc.) might be used to collect data at 120 different moments; this data serves as the input to the LSM network branch. The output dimension is the crack propagation rate, indicating that after processing and learning the input 120 time steps × 6 sensor channel data, this network branch outputs a value representing the crack propagation rate, which can be used to predict or analyze crack propagation.

[0146] Finally, the response strategy formulation unit, based on the deformation risk level, uses finite element analysis software (such as ANSYS, ABAQUS, etc.) to establish a numerical model according to the actual dimensions, material properties, and boundary conditions of the tunnel. Through finite element simulation, the unit simulates the stress and deformation of the tunnel under different working conditions, predicting potential stress concentration areas and potential failure points. Based on the constructed vibration impact database, the unit comprehensively analyzes the finite element simulation results and the data in the vibration impact database to assess the damage risk of the tunnel structure, quantifying the risk level into specific damage indicators, such as crack width and void volume, to facilitate the formulation of specific response strategies. Based on the risk assessment results, preventative measures are formulated, such as strengthening the tunnel lining and adjusting construction methods. For tunnels that have already deformed, control measures are formulated, such as installing support structures and carrying out local repairs. A long-term monitoring mechanism is established to track the tunnel's deformation in real time and adjust response strategies based on monitoring data.

[0147] In a specific application: Suppose that during the operation of a tunnel, the monitoring system detects a certain degree of voids and cracks in the tunnel lining. The response strategy formulation unit first assesses the tunnel's deformation risk level as "medium" based on data collected by sensors. Then, using the finite element method, it simulates the stress and deformation of the tunnel under the current load and geological conditions, discovering stress concentration in certain areas. Simultaneously, it extracts historical vibration data of tunnels under similar geological conditions from a vibration impact database to analyze the long-term impact of vibration on the tunnel structure. Based on these analytical results, the response strategy formulation unit assesses the damage risk data of the tunnel structure and proposes the following response strategies: adding steel mesh and shotcrete in stress concentration areas to improve the lining's crack resistance; installing vibration monitoring equipment inside the tunnel to monitor the impact of vibration in real time and adjust reinforcement measures according to vibration data; and grouting repairs in areas where cracks have already appeared to prevent further crack propagation. Through these measures, the deformation of lining voids can be effectively prevented and controlled, ensuring the structural safety of the tunnel.

[0148] The beneficial effects of the above technical solution are as follows: By adopting the solution provided in this embodiment, through intelligent monitoring and analysis methods, the deformation trend of the tunnel lining voids in strong earthquake zones is accurately tracked and monitored, providing a strong guarantee for the structural safety of the tunnel.

[0149] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A tunnel intense seismic zone lining cavity scanning system of a three-dimensional laser radar, characterized in that, include: The tunnel contour point cloud data acquisition module is used to collect and acquire tunnel contour point cloud data based on the configured autonomous navigation robot, measuring instruments and 3D LiDAR; The collaborative positioning data acquisition module is used to acquire collaborative positioning data based on a rotation speed sensor and a laser displacement sensor. The data fusion module is used to fuse multi-period tunnel contour point cloud data and collaborative positioning data based on a relative positioning algorithm to obtain fused data; the fused data refers to the dataset obtained by associating the multi-period tunnel contour point cloud data with the collaborative positioning data processed by the data fusion module. The standard tunnel profile model building module is used to process fused data using radial basis neural networks and clustering algorithms to establish a standard tunnel profile model. The lining void detection module is used to identify and detect lining voids based on a standard tunnel profile model and measured tunnel profile point cloud data.

2. A tunnel strong seismic belt lining cavity scanning system of a three-dimensional laser radar according to claim 1, characterized in that, The tunnel outline point cloud data acquisition module includes a data acquisition equipment configuration unit and a data acquisition implementation unit; The data acquisition equipment configuration unit is used to configure the autonomous navigation robot, measuring instruments, and 3D LiDAR. The autonomous navigation robot is configured to use the SLAM algorithm for trackless navigation in the tunnel and to achieve localization through fusion of the 3D LiDAR and IMU. The measuring instruments include an acoustic flaw detector and an infrared thermal imager. The 3D LiDAR is configured to perform scanning in a pulse scanning mode according to the set scanning accuracy and maximum scanning frequency. The data acquisition implementation unit is used to acquire tunnel contour point cloud data based on the configured autonomous navigation robot, measuring instruments and 3D LiDAR.

3. A three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones according to claim 2, characterized in that, Based on the configured autonomous navigation robot, measuring instruments, and 3D LiDAR, tunnel contour point cloud data is acquired, including: The autonomous navigation robot is controlled to move along a preset path inside the tunnel, while a 3D LiDAR is activated to scan and obtain 3D LiDAR point cloud data. Data on internal cracks in tunnel lining were collected using an acoustic flaw detector. Infrared thermal imagers were used to collect data on the seepage areas in the tunnel. By combining 3D lidar point cloud data, internal crack data of tunnel lining, and data of water seepage areas in the tunnel, the tunnel outline point cloud data is obtained.

4. A three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones according to claim 3, characterized in that, Controlling an autonomous navigation robot to move along a preset path within a tunnel includes: constructing the preset path based on an improved A* algorithm, where the cost function of the improved A* algorithm is defined as: ， This represents the actual path cost from the starting point to node n. This represents the heuristically estimated cost from the current node n to the destination. The deformation risk coefficient representing node n; This represents the energy consumption cost at node n; , These represent the weighting coefficients for deformation risk and energy consumption cost, respectively, and are automatically adjusted based on task priority. Among them, the deformation risk coefficient The risk contribution coefficient is obtained by weighted averaging of the risk contributions of all deformation points within the tunnel. Risk contribution is defined as the product of the reciprocal of the distance between a deformation point and the current node n, and the risk level of that deformation point. Deformation points that are closer to each other and have higher risk levels contribute more to the risk of the current node n. The specific calculation formula is as follows: ， in, This represents the total number of deformation points detected inside the tunnel. The deformation at the i-th deformation point, Represents the safety threshold of deformation. Deformation point The degree of deformation relative to the safety threshold of deformation; Deformation point The risk to node n decreases with distance; Represents the deformation point; Represents node n and deformation point Euclidean distance This represents the risk attenuation radius, used to control the scope of risk impact.

5. A three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones according to claim 1, characterized in that, Based on a rotation speed sensor and a laser displacement sensor, collaborative positioning data is acquired, including: Based on the rotation speed sensor, the kilometer marker data of the tunnel is obtained; Based on laser displacement sensors, track characteristic data of the tunnel is obtained; Based on kilometer marker data and track feature data, collaborative positioning data is generated.

6. A three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones according to claim 5, characterized in that, Based on a relative positioning algorithm, multi-period tunnel contour point cloud data and cooperative positioning data are fused to obtain fused data. The fused data refers to the dataset obtained by associating the multi-period tunnel contour point cloud data with the cooperative positioning data processed by the data fusion module. include: The formula for correcting mileage errors using kilometer marker data is as follows: ， In the above formula, Represents the error compensation coefficient. This represents the corrected mileage error. This represents the preset rotational speed of the equipment under ideal conditions. This represents the actual rotational speed measured in real time by the speed sensor. Represents the original mileage. Represents the baseline mileage; Optimize coordinate positioning using trajectory feature data from a laser displacement sensor; ， in, Let be the measurement deviation correction coefficient of the laser displacement sensor at time j, representing the contribution weight of the j-th laser ranging point to the final calibration; Let be the distance to the tunnel wall measured by the laser displacement sensor at the j-th sampling point. The pre-acquired standard geometric parameters of the track serve as the calibration reference; This represents the coordinates after coarse calibration. Coarse calibration is used to resolve large-scale mileage accumulation errors and relies on the speed sensor and the reference mileage. This represents the coordinates after fine calibration, which is used to eliminate local geometric deviations and relies on laser displacement sensors and orbital feature data.

7. A three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones according to claim 1, characterized in that, The standard tunnel outline model construction module includes a tunnel ordinary inner wall sub-model construction unit, a tunnel pipeline area sub-model construction unit, and a model summary unit; The tunnel ordinary inner wall sub-model building unit is used to perform multiple training on fused data based on radial basis function neural network, and remove large error data to establish a tunnel ordinary inner wall sub-model; The tunnel pipeline area sub-model building unit is used to train the discarded large error data based on the DBSCAN clustering algorithm to establish the tunnel pipeline area sub-model. The model aggregation unit is used to aggregate the ordinary inner wall sub-model of the tunnel and the pipeline area sub-model of the tunnel to obtain the standard tunnel outline model.

8. A three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones according to claim 7, characterized in that, Multiple training processes are performed on the fused data based on radial basis function neural networks, including: The fused data is normalized and divided into training and testing datasets; The training dataset is input into the set radial basis neural network, and the output obtains the standard contour parameters of the tunnel inner wall; During training, the prediction residual for each data point is calculated in real time, and data with prediction residuals greater than the set dynamic residual threshold are dynamically removed.

9. A three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones according to claim 7, characterized in that, The lining void detection module includes a detection report generation unit, a detection data storage unit, and a detection result display unit; The inspection report generation unit is used to input the acquired measured tunnel contour point cloud data into a standard tunnel contour model, obtain the lining void detection results, and generate an inspection report; the lining void detection results include macro void detection results and micro crack detection results; The detection data storage unit is used to store the lining void detection process data and lining void detection results based on distributed encrypted storage technology, so as to realize distributed storage and access management. The detection result display unit is used to generate a three-dimensional heat map of the lining void distribution based on the detection results of the lining voids, and to display the actual tunnel scene and the location of the lining voids.

10. A three-dimensional lidar scanning system for tunnel lining cavities in strong earthquake zones according to claim 9, characterized in that, It also includes a lining void deformation tracking and monitoring module, which is used to track and monitor the deformation trend of the lining void based on the lining void detection results and combined with external vibration monitoring data, and generate the deformation risk level of the lining void and the corresponding response strategy based on the tracking and monitoring results. The lining void deformation tracking and monitoring module includes a monitoring data acquisition unit, a deformation risk level determination unit, and a response strategy formulation unit; The monitoring data acquisition unit is used to collect deformation monitoring data of microcracks based on fiber optic grating sensor arrays and ultrasonic flaw detectors. Based on three-dimensional lidar and infrared thermal imager, deformation monitoring data of macroscopic cavities are collected; Based on a triaxial accelerometer and a ground acoustic sensor, external vibration data of the tunnel area is collected. The deformation monitoring data of microcracks, the deformation monitoring data of macrocavities, and the external vibration data of the tunnel area are summarized to obtain monitoring data; The deformation risk level determination unit is used to analyze monitoring data based on a hybrid deep learning model and output the deformation risk level of the tunnel. Specifically, it uses the convolutional neural network branch in the hybrid deep learning model to analyze the deformation monitoring data of macroscopic cavities to obtain the probability of macroscopic cavity collapse and the impact range on tunnel operation. The input data dimension of the convolutional neural network branch is 200x200x3, and the output data dimension is the collapse probability scalar. It uses the long short-term memory network branch in the hybrid deep learning model to analyze the deformation monitoring data of microscopic cracks and the external vibration data of the tunnel area to predict the microscopic crack propagation trend in the future within a set period. The input data dimension of the Long Short-Term Memory Network branch is 120 time steps × 6 sensor channels, and the output dimension is the crack propagation speed. By using the fusion layer in the hybrid deep learning model and dynamically weighting through the attention mechanism, the deformation risk level of the output tunnel is obtained based on the macroscopic cavity collapse probability and its impact range on tunnel operation, as well as the microscopic crack propagation trend in the future set period. The response strategy formulation unit is used to calculate the stress distribution of the tunnel under different vibration frequencies using the finite element simulation method, and to build a vibration impact database. By combining external vibration data of the tunnel area and using the vibration impact database, the risk data of tunnel structural damage is obtained. Based on the vibration impact database and tunnel structure damage risk data, the deformation risk level is weighted to obtain the optimized deformation risk level. Develop matching response strategies based on the optimized deformation risk level.

Images